#3 - Dario Amodei on OpenAI and how AI will change the world for good and ill

Today's episode is a compilation of interviews I recently recorded for two other shows, Love Your Work and The Neoliberal Podcast.  If you've listened to absolutely everything on this podcast feed, you'll have heard four interviews with me already, but fortunately I don't think these two include much repetition, and I've gotten a decent amount of positive feedback on both.  First up, I speak with David Kadavy on his show, Love Your Work.  This is a particularly personal and relaxed interview. We talk about all sorts of things, including nicotine gum, plastic straw bans, whether recycling is important, how many lives a doctor saves, why interviews should go for at least 2 hours, how athletes doping could be good for the world, and many other fun topics.  • Our annual impact survey is about to close — I'd really appreciate if you could take 3–10 minutes to fill it out now.  • The blog post about this episode. At some points we even actually discuss effective altruism and 80,000 Hours, but you can easily skip through those bits if they feel too familiar.  The second interview is with Jeremiah Johnson on the Neoliberal Podcast. It starts 2 hours and 15 minutes into this recording.  Neoliberalism in the sense used by this show is not the free market fundamentalism you might associate with the term. Rather it's a centrist or even centre-left view that supports things like social liberalism, multilateral international institutions, trade, high rates of migration, racial justice, inclusive institutions, financial redistribution, prioritising the global poor, market urbanism, and environmental sustainability.  This is the more demanding of the two conversations, as listeners to that show have already heard of effective altruism, so we were able to get the best arguments Jeremiah could offer against focusing on improving the long term future of the world.  Jeremiah is more of a fan of donating to evidence-backed global health charities recommended by GiveWell, and does so himself.  I appreciate him having done his homework and forcing me to do my best to explain how well my views can stand up to counterarguments. It was a challenge for me to paint the whole picture in the half an hour we spent on longterm and I expect there's answers in there which will be fresh even for regular listeners.  I hope you enjoy both conversations! Feel free to email me with any feedback. The 80,000 Hours Podcast is produced by Keiran Harris.

25 Syys 20193h 14min

1. Fill out our annual impact survey here. 2. Find a great vacancy on our job board. 3. Learn about our key ideas, and get links to our top articles. 4. Join our newsletter for an email about what's new, every 2 weeks or so. 5. Or follow our pages on Facebook and Twitter. —— Once a year 80,000 Hours runs a survey to find out whether we've helped our users have a larger social impact with their life and career. We and our donors need to know whether our services, like this podcast, are helping people enough to continue them or scale them up, and it's only by hearing from you that we can make these decisions in a sensible way. So, if 80,000 Hours' podcast, job board, articles, headhunting, advising or other projects have somehow contributed to your life or career plans, please take 3–10 minutes to let us know how. You can also let us know where we've fallen short, which helps us fix problems with what we're doing. We've refreshed the survey this year, hopefully making it easier to fill out than in the past. We'll keep this appeal up for about two weeks, but if you fill it out now that means you definitely won't forget! Thanks so much, and talk to you again in a normal episode soon. — RobTag for internal use: this RSS feed is originating in BackTracks.

16 Syys 20193min

Historically, progress in the field of cryptography has had major consequences. It has changed the course of major wars, made it possible to do business on the internet, and enabled private communication between both law-abiding citizens and dangerous criminals. Could it have similarly significant consequences in future? Today's guest — Vitalik Buterin — is world-famous as the lead developer of Ethereum, a successor to the cryptographic-currency Bitcoin, which added the capacity for smart contracts and decentralised organisations. Buterin first proposed Ethereum at the age of 20, and by the age of 23 its success had likely made him a billionaire. At the same time, far from indulging hype about these so-called 'blockchain' technologies, he has been candid about the limited good accomplished by Bitcoin and other currencies developed using cryptographic tools — and the breakthroughs that will be needed before they can have a meaningful social impact. In his own words, *"blockchains as they currently exist are in many ways a joke, right?"* But Buterin is not just a realist. He's also an idealist, who has been helping to advance big ideas for new social institutions that might help people better coordinate to pursue their shared goals. Links to learn more, summary and full transcript. By combining theories in economics and mechanism design with advances in cryptography, he has been pioneering the new interdiscriplinary field of 'cryptoeconomics'. Economist Tyler Cowen hasobserved that, "at 25, Vitalik appears to repeatedly rediscover important economics results from famous papers, without knowing about the papers at all." Along with previous guest Glen Weyl, Buterin has helped develop a model for so-called 'quadratic funding', which in principle could transform the provision of 'public goods'. That is, goods that people benefit from whether they help pay for them or not. Examples of goods that are fully or partially 'public goods' include sound decision-making in government, international peace, scientific advances, disease control, the existence of smart journalism, preventing climate change, deflecting asteroids headed to Earth, and the elimination of suffering. Their underprovision in part reflects the difficulty of getting people to pay for anything when they can instead free-ride on the efforts of others. Anything that could reduce this failure of coordination might transform the world. But these and other related proposals face major hurdles. They're vulnerable to collusion, might be used to fund scams, and remain untested at a small scale — not to mention that anything with a square root sign in it is going to struggle to achieve societal legitimacy. Is the prize large enough to justify efforts to overcome these challenges? In today's extensive three-hour interview, Buterin and I cover: • What the blockchain has accomplished so far, and what it might achieve in the next decade; • Why many social problems can be viewed as a coordination failure to provide a public good; • Whether any of the ideas for decentralised social systems emerging from the blockchain community could really work; • His view of 'effective altruism' and 'long-termism'; • Why he is optimistic about 'quadratic funding', but pessimistic about replacing existing voting with 'quadratic voting'; • Why humanity might have to abandon living in cities; • And much more. Get this episode by subscribing to our podcast on the world’s most pressing problems and how to solve them: type 80,000 Hours into your podcasting app. The 80,000 Hours Podcast is produced by Keiran Harris.

3 Syys 20193h 18min

Imagine that – one day – humanity dies out. At some point, many millions of years later, intelligent life might well evolve again. Is there any message we could leave that would reliably help them out? In his second appearance on the 80,000 Hours Podcast, machine learning researcher and polymath Paul Christiano suggests we try to answer this question with a related thought experiment: are there any messages we might want to send back to our ancestors in the year 1700 that would have made history likely to go in a better direction than it did? It seems there probably are. • Links to learn more, summary, and full transcript. • Paul's first appearance on the show in episode 44. • An out-take on decision theory. We could tell them hard-won lessons from history; mention some research questions we wish we'd started addressing earlier; hand over all the social science we have that fosters peace and cooperation; and at the same time steer clear of engineering hints that would speed up the development of dangerous weapons. But, as Christiano points out, even if we could satisfactorily figure out what we'd like to be able to tell our ancestors, that's just the first challenge. We'd need to leave the message somewhere that they could identify and dig up. While there are some promising options, this turns out to be remarkably hard to do, as anything we put on the Earth's surface quickly gets buried far underground. But even if we figure out a satisfactory message, and a ways to ensure it's found, a civilization this far in the future won't speak any language like our own. And being another species, they presumably won't share as many fundamental concepts with us as humans from 1700. If we knew a way to leave them thousands of books and pictures in a material that wouldn't break down, would they be able to decipher what we meant to tell them, or would it simply remain a mystery? That's just one of many playful questions discussed in today's episode with Christiano — a frequent writer who's willing to brave questions that others find too strange or hard to grapple with. We also talk about why divesting a little bit from harmful companies might be more useful than I'd been thinking. Or whether creatine might make us a bit smarter, and carbon dioxide filled conference rooms make us a lot stupider. Finally, we get a big update on progress in machine learning and efforts to make sure it's reliably aligned with our goals, which is Paul's main research project. He responds to the views that DeepMind's Pushmeet Kohli espoused in a previous episode, and we discuss whether we'd be better off if AI progress turned out to be most limited by algorithmic insights, or by our ability to manufacture enough computer processors. Some other issues that come up along the way include: • Are there any supplements people can take that make them think better? • What implications do our views on meta-ethics have for aligning AI with our goals? • Is there much of a risk that the future will contain anything optimised for causing harm? • An out-take about the implications of decision theory, which we decided was too confusing and confused to stay in the main recording. Get this episode by subscribing to our podcast on the world’s most pressing problems and how to solve them: type 80,000 Hours into your podcasting app. Or read the transcript below. The 80,000 Hours Podcast is produced by Keiran Harris.

5 Elo 20192h 11min

From 1870 to 1950, the introduction of electricity transformed life in the US and UK, as people gained access to lighting, radio and a wide range of household appliances for the first time. Electricity turned out to be a general purpose technology that could help with almost everything people did. Some think this is the best historical analogy we have for how machine learning could alter life in the 21st century. In addition to massively changing everyday life, past general purpose technologies have also changed the nature of war. For example, when electricity was introduced to the battlefield, commanders gained the ability to communicate quickly with units in the field over great distances. How might international security be altered if the impact of machine learning reaches a similar scope to that of electricity? Today's guest — Helen Toner — recently helped found the Center for Security and Emerging Technology at Georgetown University to help policymakers prepare for such disruptive technical changes that might threaten international peace. • Links to learn more, summary and full transcript • Philosophy is one of the hardest grad programs. Is it worth it, if you want to use ideas to change the world? by Arden Koehler and Will MacAskill • The case for building expertise to work on US AI policy, and how to do it by Niel Bowerman • AI strategy and governance roles on the job board Their first focus is machine learning (ML), a technology which allows computers to recognise patterns, learn from them, and develop 'intuitions' that inform their judgement about future cases. This is something humans do constantly, whether we're playing tennis, reading someone's face, diagnosing a patient, or figuring out which business ideas are likely to succeed. Sometimes these ML algorithms can seem uncannily insightful, and they're only getting better over time. Ultimately a wide range of different ML algorithms could end up helping us with all kinds of decisions, just as electricity wakes us up, makes us coffee, and brushes our teeth -- all in the first five minutes of our day. Rapid advances in ML, and the many prospective military applications, have people worrying about an 'AI arms race' between the US and China. Henry Kissinger and the past CEO of Google Eric Schmidt recently wrote that AI could "destabilize everything from nuclear détente to human friendships." Some politicians talk of classifying and restricting access to ML algorithms, lest they fall into the wrong hands. But if electricity is the best analogy, you could reasonably ask — was there an arms race in electricity in the 19th century? Would that have made any sense? And could someone have changed the course of history by changing who first got electricity and how they used it, or is that a fantasy? In today's episode we discuss the research frontier in the emerging field of AI policy and governance, how to have a career shaping US government policy, and Helen's experience living and studying in China. We cover: • Why immigration is the main policy area that should be affected by AI advances today. • Why talking about an 'arms race' in AI is premature. • How Bobby Kennedy may have positively affected the Cuban Missile Crisis. • Whether it's possible to become a China expert and still get a security clearance. • Can access to ML algorithms be restricted, or is that just not practical? • Whether AI could help stabilise authoritarian regimes. Get this episode by subscribing to our podcast on the world’s most pressing problems and how to solve them: type 80,000 Hours into your podcasting app. The 80,000 Hours Podcast is produced by Keiran Harris.

17 Heinä 20191h 54min

Have you ever been infuriated by a doctor's unwillingness to give you an honest, probabilistic estimate about what to expect? Or a lawyer who won't tell you the chances you'll win your case? Their behaviour is so frustrating because accurately predicting the future is central to every action we take. If we can't assess the likelihood of different outcomes we're in a complete bind, whether the decision concerns war and peace, work and study, or Black Mirror and RuPaul's Drag Race. Which is why the research of Professor Philip Tetlock is relevant for all of us each and every day. He has spent 40 years as a meticulous social scientist, collecting millions of predictions from tens of thousands of people, in order to figure out how good humans really are at foreseeing the future, and what habits of thought allow us to do better. Along with other psychologists, he identified that many ordinary people are attracted to a 'folk probability' that draws just three distinctions — 'impossible', 'possible' and 'certain' — and which leads to major systemic mistakes. But with the right mindset and training we can become capable of accurately discriminating between differences as fine as 56% as against 57% likely. • Links to learn more, summary and full transcript • The calibration training app • Sign up for the Civ-5 counterfactual forecasting tournament • A review of the evidence on good forecasting practices • Learn more about Effective Altruism Global In the aftermath of Iraq and WMDs the US intelligence community hired him to prevent the same ever happening again, and his guide — Superforecasting: The Art and Science of Prediction — became a bestseller back in 2014. That was five years ago. In today's interview, Tetlock explains how his research agenda continues to advance, today using the game Civilization 5 to see how well we can predict what would have happened in elusive counterfactual worlds we never get to see, and discovering how simple algorithms can complement or substitute for human judgement. We discuss how his work can be applied to your personal life to answer high-stakes questions, like how likely you are to thrive in a given career path, or whether your business idea will be a billion-dollar unicorn — or fall apart catastrophically. (To help you get better at figuring those things out, our site now has a training app developed by the Open Philanthropy Project and Clearer Thinking that teaches you to distinguish your '70 percents' from your '80 percents'.) We also bring some tough methodological questions raised by the author of a recent review of the forecasting literature. And we find out what jobs people can take to make improving the reasonableness of decision-making in major institutions that shape the world their profession, as it has been for Tetlock over many decades. We view Tetlock's work as so core to living well that we've brought him back for a second and longer appearance on the show — his first was back in episode 15. Get this episode by subscribing to our podcast on the world’s most pressing problems and how to solve them: type 80,000 Hours into your podcasting app. The 80,000 Hours Podcast is produced by Keiran Harris.

28 Kesä 20192h 11min

It can often feel hopeless to be an activist seeking social change on an obscure issue where most people seem opposed or at best indifferent to you. But according to a new book by Professor Cass Sunstein, they shouldn't despair. Large social changes are often abrupt and unexpected, arising in an environment of seeming public opposition.The Communist Revolution in Russia spread so swiftly it confounded even Lenin. Seventy years later the Soviet Union collapsed just as quickly and unpredictably.In the modern era we have gay marriage, #metoo and the Arab Spring, as well as nativism, Euroskepticism and Hindu nationalism.How can a society that so recently seemed to support the status quo bring about change in years, months, or even weeks?Sunstein — coauthor of Nudge, Obama White House official, and by far the most cited legal scholar of the late 2000s — aims to unravel the mystery and figure out the implications in his new book How Change Happens. He pulls together three phenomena which social scientists have studied in recent decades: preference falsification, variable thresholds for action, and group polarisation. If Sunstein is to be believed, together these are a cocktail for social shifts that are chaotic and fundamentally unpredictable. • Links to learn more, summary and full transcript. • 80,000 Hours Annual Review 2018. • How to donate to 80,000 Hours. In brief, people constantly misrepresent their true views, even to close friends and family. They themselves aren't quite sure how socially acceptable their feelings would have to become, before they revealed them, or joined a campaign for social change. And a chance meeting between a few strangers can be the spark that radicalises a handful of people, who then find a message that can spread their views to millions. According to Sunstein, it's "much, much easier" to create social change when large numbers of people secretly or latently agree with you. But 'preference falsification' is so pervasive that it's no simple matter to figure out when that's the case. In today's interview, we debate with Sunstein whether this model of cultural change is accurate, and if so, what lessons it has for those who would like to shift the world in a more humane direction. We discuss: • How much people misrepresent their views in democratic countries. • Whether the finding that groups with an existing view tend towards a more extreme position would stand up in the replication crisis. • When is it justified to encourage your own group to polarise? • Sunstein's difficult experiences as a pioneer of animal rights law. • Whether activists can do better by spending half their resources on public opinion surveys. • Should people be more or less outspoken about their true views? • What might be the next social revolution to take off? • How can we learn about social movements that failed and disappeared? • How to find out what people really think. Chapters:• Rob’s intro (00:00:00)• Cass's Harvard lecture on How Change Happens (00:02:59)• Rob & Cass's conversation about the book (00:41:43) The 80,000 Hours Podcast is produced by Keiran Harris.

17 Kesä 20191h 43min

When you're building a bridge, responsibility for making sure it won't fall over isn't handed over to a few 'bridge not falling down engineers'. Making sure a bridge is safe to use and remains standing in a storm is completely central to the design, and indeed the entire project.When it comes to artificial intelligence, commentators often distinguish between enhancing the capabilities of machine learning systems and enhancing their safety. But to Pushmeet Kohli, principal scientist and research team leader at DeepMind, research to make AI robust and reliable is no more a side-project in AI design than keeping a bridge standing is a side-project in bridge design.Far from being an overhead on the 'real' work, it’s an essential part of making AI systems work at all. We don’t want AI systems to be out of alignment with our intentions, and that consideration must arise throughout their development.Professor Stuart Russell — co-author of the most popular AI textbook — has gone as far as to suggest that if this view is right, it may be time to retire the term ‘AI safety research’ altogether. • Want to be notified about high-impact opportunities to help ensure AI remains safe and beneficial? Tell us a bit about yourself and we’ll get in touch if an opportunity matches your background and interests. • Links to learn more, summary and full transcript. • And a few added thoughts on non-research roles. With the goal of designing systems that are reliably consistent with desired specifications, DeepMind have recently published work on important technical challenges for the machine learning community. For instance, Pushmeet is looking for efficient ways to test whether a system conforms to the desired specifications, even in peculiar situations, by creating an 'adversary' that proactively seeks out the worst failures possible. If the adversary can efficiently identify the worst-case input for a given model, DeepMind can catch rare failure cases before deploying a model in the real world. In the future single mistakes by autonomous systems may have very large consequences, which will make even small failure probabilities unacceptable. He's also looking into 'training specification-consistent models' and formal verification', while other researchers at DeepMind working on their AI safety agenda are figuring out how to understand agent incentives, avoid side-effects, and model AI rewards. In today’s interview, we focus on the convergence between broader AI research and robustness, as well as: • DeepMind’s work on the protein folding problem • Parallels between ML problems and past challenges in software development and computer security • How can you analyse the thinking of a neural network? • Unique challenges faced by DeepMind’s technical AGI safety team • How do you communicate with a non-human intelligence? • What are the biggest misunderstandings about AI safety and reliability? • Are there actually a lot of disagreements within the field? • The difficulty of forecasting AI development Get this episode by subscribing to our podcast on the world’s most pressing problems and how to solve them: type 80,000 Hours into your podcasting app. Or read the transcript below. The 80,000 Hours Podcast is produced by Keiran Harris.

3 Kesä 20191h 30min